Seven tests. Mostly null.
Published as tested.
Three generations of pre-registered MES momentum and exit-law tests, run directly for this site, plus a sibling registry's multi-instrument trial runs held to the same one-shot-holdout standard. Every result below is published as it came back, including the ones that did not work.
The three-generation arc
Pre-registered momentum and exit-law tests on MES micro futures, each benchmarked against a coin-flip twin sharing the identical bracket.
19 sessions looked real. 488 said no.
A fixed-threshold momentum entry looked promising on a 19-session calibration slice. A 488-session out-of-sample replay found no edge.
We fixed the confound. The result barely moved.
Fixing the entry law for price drift did not fix the result: real +0.297 pts/trade on a 123-session holdout, paired t=0.55, against a pre-declared bar of t=2.0.
A convex exit made money on MES. So did a coin flip, to the decimal.
A pre-registered, one-shot holdout test of a momentum exit law on MES vs. a coin-flip twin sharing the identical bracket and stop logic. Paired t=-0.01.
The winnow sibling registry
A separate organism's pre-registered trial program across three instruments, held to the same one-shot-holdout standard as our own tests.
7,000 pre-registered trials. One promotion. Zero survivors.
A sibling registry logged 7,000 pre-registered MES trials. One cleared its own bar, was halted by standing order, and failed its sealed one-shot holdout.
1,100 gold trials. One promotion. Zero survivors.
1,100 pre-registered gold trials across 11 strategy families. One cleared its bar; a standing cost floor killed it before the holdout.
2,200 crude oil trials. Zero promotions.
2,200 pre-registered crude oil trials across 11 strategy families. Not one candidate reached the statistical bar.
A new indicator. 1,600 trials. Zero survivors.
A new indicator, tested 1,600 times across eight strategy families on the same instrument. Nothing crossed the bar.